National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The...

14
National Gallery Technical Bulletin Volume 23, 2002 National Gallery Company London Distributed by Yale University Press

Transcript of National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The...

Page 1: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

NationalGallery Technical BulletinVolume 23, 2002

National Gallery CompanyLondon

Distributed by Yale University Press

Page 2: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

Series editor Ashok Roy

© National Gallery Company Limited 2002

All rights reserved. No part of this publication may betransmitted in any form or by any means, electronic ormechanical, including photocopy, recording, or any infor-mation storage and retrieval system, without the priorpermission in writing of the publisher.

First published in Great Britain in 2002 by National Gallery Company LimitedSt Vincent House, 30 Orange StreetLondon wc2h 7hh

www.nationalgallery.co.uk

British Library Cataloguing in Publication DataA catalogue record for this journal is available from the British Library

isbn 1 85709 941 9

issn 0140 7430

525039

Edited by Diana DaviesProject manager Jan GreenDesign by Tim HarveyPrinted in Italy by Conti Tipocolor

FRONT COVER

Garofalo, Saint Augustine with the Holy Family and SaintCatherine of Alexandria (NG 81), (detail of plate 4, p. 23).

TITLE PAGE

Garofalo, The Virgin and Child with Saints William ofAquitaine, Clare (?), Anthony of Padua and Francis(NG 671), (detail of plate 3, p. 22).

This volume of the Technical Bulletin is published withthe generous support of the Samuel H. Kress Foundation.

Page 3: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

Introduction

The darkening of the red mercury sulphidepigment vermilion is a well-known phenome-

non, but one which is not fully understood. Onsome paintings vermilion has survived for hundredsof years unchanged, while on others it has devel-oped a black or silver-grey crust that can beextremely disfiguring. Some insight into the processhas been gained from previous studies, which haveestablished from experiments on test samples thatvarious impurities can accelerate the blackening.1

Vermilion made by the wet process, discovered in1687, can contain impurities introduced duringmanufacture.2 The dry-process vermilion andcinnabar used in earlier paintings is generally verypure, but it can still exhibit severe blackening.3 Thispaper describes our investigation of the phenome-non through the detailed analysis of samples ofdiscoloured vermilion from paintings dating frombefore 1687. Ten paintings in the National Galleryand two paintings from the Courtauld Gallery werestudied (Table 1). They range in date from the four-teenth to the seventeenth century, and were allpainted before the introduction of wet-processvermilion, so dry-process vermilion or mineralcinnabar must have been used.

Examination and analysis

The paintings examined illustrate well the unpre-dictable nature of the blackening of vermilion. Insome paintings, such as Jacob Jordaens’ HolyFamily and Saint John the Baptist (NG 164), only afew brushstrokes of vermilion have deteriorated.The Virgin’s cloak in Jordaens’ painting is paintedwith red lake and vermilion, and some of the finalstrokes of vermilion, probably originally highlights,have become completely grey (plates 1 and 2).There are many areas of blackened vermilion inUccello’s Battle of San Romano (NG 583), such asthe shafts of some of the broken lances, the legs ofsome of the figures in the foreground, and the

50 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23

The Blackening of Vermilion: An Analytical Study ofthe Process in Paintings

marika spring and rachel grout

Table 1. List of paintings examined

• Bernardo Daddi, The Crucifixion and Saints(Courtauld Gallery, Gambier-Parry bequestchecklist no.82), 1348, egg tempera onpanel.

• Nardo di Cione, Saint John the Baptist withSaint John the Evangelist(?) and Saint James(NG 581), c.1365, egg tempera on poplar.

• English School, The Estouteville Triptych(Courtauld Gallery, cat.no.109), c.1360–7,oil on panel.

• Attributed to Jacopo di Cione andWorkshop, San Pier Maggiore Altarpiece(NG 569–578), 1370–1, egg tempera onpoplar.

• Attributed to Niccolò di Pietro Gerini,Triptych: The Baptism of Christ (NG579.1–5), probably 1387, egg tempera onpanel.

• Lorenzo Monaco, The Coronation of theVirgin (NG 1897), probably 1407–9, eggtempera on poplar.

• Giovanni del Ponte, The Ascension of SaintJohn the Evangelist, with Saints (NG 580),probably 1410–20, egg tempera on panel.

• Paolo Uccello, The Battle of San Romano(NG 583), probably 1450–60, egg temperaon panel.

• Benozzo Gozzoli, The Virgin and ChildEnthroned among Angels and Saints (NG283), 1461–2, egg tempera on panel.

• Workshop of the Master of the FemaleHalf-lengths, Saint John on Patmos (NG717), c.1525–50, oil on oak.

• Jacob Jordaens, The Holy Family and SaintJohn the Baptist (NG 164), probably1620–5, oil on oak.

• Aelbert Cuyp, A Hilly River Landscape witha Horseman talking to a Shepherdess (NG53), 1655–60, oil on canvas.

Page 4: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

saddlery and harness of the white horse.4 A borderaround the edge of the harness has remained red,however, as have some other seemingly randompatches (plate 3). On many of the paintings exam-ined only part of the modelling of a red drapery inwhich vermilion has been used has deteriorated. Forexample, in Benozzo Gozzoli’s Virgin and ChildEnthroned (NG 283), the highlights of the red drap-ery around the Christ Child have becomecompletely grey while the shadows are still red(plate 4). Something similar can be seen on TheAscension of Saint John the Evangelist, with Saints(NG 580) by Giovanni del Ponte, where the blacken-ing on Saint Jerome’s red robe is not uniform, butfollows the modelling (plate 5). Individual black-ened strokes of paint can be seen next to strokes ofred paint that have not deteriorated.

The appearance of the darkened paint surfacesalso varies. On many of the paintings it is greyrather than black, and can have a lilac-grey colourwhere the deteriorated crust is thin; this is becausethe unchanged red vermilion below is contributingto the colour at the surface, as on the painting byBenozzo Gozzoli. In other paintings, such asGiovanni del Ponte’s altarpiece of Saint John theEvangelist, it is a much darker grey colour.

The reason for the variation in colour becameclearer when examining the surface of samples fromdiscoloured areas under the microscope. plate 6

shows the surface of the sample from the paintingby Benozzo Gozzoli. A crust of deterioratedpigment that consists of white, as well as black,particles has formed on the surface of the paint,beneath the varnish. Focusing through the thicknessof the crust with the microscope, very thin insub-stantial islands of black material can be seen lyingon the surface of unchanged red vermilion, beneathbulkier white, or greyish-white, particles.

The deteriorated crust consists of both blackand white material in all the paintings with dark-ened vermilion that were examined, a significantobservation that has not been made in earlier stud-ies. The appearance of the various samples underthe microscope was remarkably consistent, differingonly in the proportion of black to white material,which correlated with the appearance of the deteri-

NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51

The Blackening of Vermilion: An Analytical Study of the Process in Paintings

plate 1 Jacob Jordaens, The Holy Family with Saint Johnthe Baptist (NG 164), probably 1620–5. Oak, 123 × 93.9cm.

plate 2 Jacob Jordaens, The Holy Family with Saint Johnthe Baptist. Detail of the Virgin’s cloak.

plate 3 Paolo Uccello, Niccolò Mauruzi da Tolentino atthe Battle of San Romano (NG 583), c.1440. Detail ofdarkened vermilion in the white charger’s harness.

Page 5: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

oration on the painting itself. In the sample fromthe painting by Benozzo, there is a high proportionof white material, explaining why the deterioratedarea appears purplish grey in colour. In a samplefrom the darker grey discoloration on the altarpieceby Giovanni del Ponte there is a higher proportionof black material in the crust although, as always,some white material is also present (plate 7). Onthe predella panel of the altarpiece of The Baptismof Christ by Niccolò di Pietro Gerini, the bedcoveris painted with vermilion, which has discoloured toa particularly light grey colour. A high proportionof white material is visible on the surface of thesample (plate 8).

The character of the black material is especiallyvisible in a sample from the blackened vermilionrobe of one of the saints on the left panel of the SanPier Maggiore Altarpiece (NG 569), attributed toJacopo di Cione and Workshop. Islands of very fine-grained black are visible on the surface ofunchanged red vermilion (plate 9). The larger blackparticles that can be seen are dirt. They lie on top ofthe white particles, so can easily be distinguishedfrom the black deterioration product. Although it isdifficult to see any white material on the surface ofthis particular sample, some white particles arepresent and are visible in a cross-section of thesample.

52 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23

Marika Spring and Rachel Grout

plate 4 Benozzo Gozzoli, The Virgin and Child Enthroned(NG 283), 1461–2. Detail of the Christ Child.

plate 5 Giovanni del Ponte, The Ascension of Saint Johnthe Evangelist, with Saints (NG 580), probably c.1410–20.Detail of Saint Jerome.

Page 6: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 53

The Blackening of Vermilion: An Analytical Study of the Process in Paintings

a. Detail of discoloured vermilion inthe drapery around the Christ Child.

b. Surface of a paint sample from thearea of discoloured vermilion.Original magnification 250×; actualmagnification 110×.

c. Cross-section of a paint sample fromthe area of discoloured vermilion. Thegrey chloride-containing crust, withunchanged red vermilion beneath it, isrelatively thick. Original magnification500×; actual magnification 450×.

plate 6 Benozzo Gozzoli, The Virgin and Child Enthroned (NG 283).

a. Detail of discoloured vermilion onSaint Jerome’s robe.

plate 8 Niccolò di Pietro Gerini, Triptych: The Baptism of Christ (NG 579.1–5).

plate 7 Giovanni del Ponte, The Ascension of Saint John the Evangelist, with Saints (NG 580).

b. Surface of a paint sample from thearea of discoloured vermilion. Originalmagnification 250×; actual magnifica-tion 160×.

c. Cross-section of a paint sample fromthe area of discoloured vermilion. Thegrey crust on the surface of the paint isextremely thin. A few black particlesare just visible beneath some whitematerial containing chloride. Originalmagnification 1250×; actual magnifica-tion 560×.

a. Detail of discoloured vermilion onthe bedcover in the predella paneldepicting the birth of Saint John theBaptist.

b. Surface of a paint sample from thediscoloured vermilion on the bedcover.Original magnification 250×; actualmagnification 160×.

c. Cross-section of a paint sample fromthe area of discoloured vermilion. Thegrey deteriorated crust on the paintsurface contains a high proportion ofwhite chloride-containing material.Original magnification 1250×; actualmagnification 740×.

Page 7: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

The cross-sections of samples from areas ofdeteriorated vermilion show clearly that the deterio-rated crust is extremely thin, even when thediscoloration is severe enough to have a strongvisual impact on the painting. In many cross-sections it is around five microns in thickness, and isbarely perceptible. Some white particles, and a fewblack particles lying beneath the white, are just visi-ble on the surface of the cross-section from thepainting by Giovanni del Ponte (plate 7c), confirm-ing what was observed in the unmounted samples.

The crust is slightly thicker in the cross-sectionofthe sample from the painting by Niccolò di PietroGerini. There is more white material in this sample,which forms a distinct rather crumbly and friablecrust on the surface of unchanged vermilion (plate

8c). In this case, as in most of the cross-sectionsexamined, the black material was too thin to beseen.

Having established that both black and whitedeterioration products are present, further analysiswas carried out in an attempt to identify them. Spotanalysis by Energy Dispersive X-ray (EDX) analysiswas carried out in the scanning electron microscope,directed at a white area. In the EDX spectrum of thesample from the painting by Benozzo Gozzoli thereis a very clear and clean peak for chlorine, as well aspeaks that can be assigned to mercury, suggestingthat a mercury chloride compound is present (fig.1). The detection of mercury as well as chlorine inthe white particles confirms that they are a deterio-ration product of the vermilion, and not simplysome random unrelated surface accretion. SimilarEDX results were obtained for all of the paintingslisted. In every case, white particles containingmercury and chlorine were found in the deterioratedcrust.

Since several mercury chloride compounds arewhite, the next step was to try to establish whichcompound was present, using Raman microscopy.This technique is dependent on the behaviour of themolecule as a whole, and thus can distinguishbetween different compounds containing the sameelements. Analysis was carried out directly on cross-sections from two of the paintings in the study,Benozzo Gozzoli’s Virgin and Child Enthroned andthe Saint John on Patmos (NG 717) attributed to theWorkshop of the Master of the Female Half-lengths, again focusing on white particles in the greycrust on the surface of the paint. From this, themercury chloride compound could be identifiedmore specifically as calomel (mercury (I) chloride).fig. 2 illustrates the Raman spectrum collected fromthe sample from the painting by Benozzo Gozzoli.The major peak in the spectrum from the sample, at164 cm-1, can be seen to correspond to the majorpeak in the standard spectrum of calomel.5 Theuppermost spectrum in fig. 2 is that of a standardsample of vermilion, and a comparison with thespectrum from the sample shows that the smallpeaks are likely to be vermilion. The resolution ofRaman microscopy is very good, but some of thevermilion immediately underneath the deterioratedcrust may have contributed to the spectrum.

54 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23

Marika Spring and Rachel Grout

fig. 1 EDX spectrum of a white particle in the deterio-rated crust on the surface of the sample from discolouredvermilion in Benozzo Gozzoli’s Virgin and ChildEnthroned.

plate 9 Attributed to Jacopo di Cione and Workshop, SanPier Maggiore Altarpiece, main tier (NG 569–578), 1370–1.Surface of a sample from the discoloured vermilion robe ofa saint in the left panel of the main tier. Original magnifi-cation 220×; actual magnification 170×.

Page 8: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

The black deterioration product has alwaysbeen thought to be metacinnabar, the black cubicform of mercury sulphide, although this has neverbeen convincingly proved. X-ray diffraction analy-sis of blackened vermilion in previous studies hasgiven a pattern for cinnabar, but with reducedintensity relative to unblackened vermilion. It hasbeen suggested that this may be because the blackmaterial is amorphous, or because the crystalitesare too small for successful X-ray diffraction analy-sis.6 With the aim of assessing the potential ofRaman microscopy for the identification ofmetacinnabar, a spectrum of a standard sample wascollected. The spectrum produced was identical tothat of red vermilion, but of reduced intensity, sounfortunately it appears that Raman microscopywill not settle the question of whether the blackdeterioration product is metacinnabar.7

No chlorine was detected by EDX analysis inthe unchanged particles of vermilion below thepaint surface in the cross-sections. Without excep-tion they were very pure, containing only a smallamount of silicon in addition to mercury andsulphur.8 EDX mapping of the cross-sectionsconfirmed that the chlorine was concentrated in thedeteriorated crust. Since the chlorine is present onlyat the surface of the sample, it must originate froman external source, the most likely being dirt. EDXanalysis of dirt collected in a filter from a dirtmonitor placed in Gallery 45 in the NationalGallery showed the presence of chlorine, togetherwith some sodium, which suggests that the chlorineis present as sodium chloride (fig. 3). The sodiumchloride probably originates from skin debris, sinceGC–MS analysis of this sample detected squalene,which is a major component of skin grease.9

Analysis of dirt in other galleries has also shownthe presence of small amounts of chloride. Gyselsand Van Grieken (1999) have carried out quantita-tive analysis on dirt collected from three majormuseums, and attempted to characterise the differ-ent types of particle present in the dirt.Calcium-rich and calcium-silicon-rich particleswere found in the dirt deposits collected, often withsodium chloride and sulphur adsorbed onto theirsurface.10 This agrees well with the results of spotanalyses of dirt collected in the National Gallery,and also explains why in some of the EDX analysesof cross-sections of discoloured vermilion, somecalcium-containing particles were detected on thesurface of the paint.11

NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 55

The Blackening of Vermilion: An Analytical Study of the Process in Paintings

fig. 2 Raman spectrum of a white particle in the grey deteriorated crust on the surface of a sample from an areaof discoloured vermilion on Benozzo Gozzoli’s Virgin andChild Enthroned, with spectra of standard samples ofcalomel (Hg

2Cl

2) and vermilion for comparison.

fig. 3 EDX spectrum of dirt collected from Gallery 45.

Vermilion

Calomel

Paint sample

Wavenumber (cm-1)

200 400

Inte

nsity

Page 9: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

In the case of one of the paintings studied,Uccello’s Battle of San Romano, it can be demon-strated that the colour change of the vermilion didnot happen recently, but occurred before the adventof air-conditioning and filtration of dirt in theNational Gallery around the middle of the twentiethcentury.12 The source of the heavy pollution forwhich London was notorious until the Clean Air Actof 1956 was coal burning. Some deposits of coalcontain high levels of chloride ions and could releasechloride into the atmosphere during combustion.13

The results of some nineteenth-century analyses ofrain water in central London, designed to measurelevels of pollution, confirm that chloride was presentin the atmosphere at this time.14

The role of chloride in the blackening of vermilion

Chloride has already been identified in previousstudies as one of the ionic species that can acceleratethe blackening of vermilion.15 Vincent Daniels foundthat sodium chloride added to vermilion tube water-colour accelerated blackening, and that the rate ofblackening increased with an increase in relativehumidity.16 A recent paper by J.K. McCormack onthe darkening of mineral specimens of cinnabar alsohighlighted the role that chloride can play in theprocess. McCormack found that cinnabar fromcertain deposits is photosensitive (that is, it blackensin response to light), and that photosensitivecinnabar contains between 0.05% and 1% chlorine,while non-photosensitive cinnabar contains less than0.01%. Furthermore, cinnabar from the historicsources of Monte Amiata (Italy) and Almadèn(Spain), which is not photosensitive, became photo-sensitive following exposure to halides.17

A possible mechanism for the deterioration ofvermilion on paintings can be proposed from theresults of McCormack’s experiments to simulate theprocess. He found that reaction of vermilion withhydrochloric acid produced the compoundcorderoite (Hg

3S2Cl

2). He also found that the same

product was produced when vermilion was left for along period of time in a solution of sodium chloride.It is possible, then, that sodium chloride in acidichygroscopic dirt particles deposited on the paintsurface, in a moist environment, could initiate thistype of reaction on paintings. Corderoite is photo-sensitive and so degrades under the action of light.X-ray diffraction of the grey photodegradationproduct of corderoite, carried out by McCormackduring his experiments, gave a pattern for calomel,an interesting result since it is the degradation prod-

uct observed on paintings.18 Although calomel is white, the photodegradation

product of the corderoite that McCormack madewas dark grey in bulk appearance, so it is likely thatsome black material is also present which was notdetectable by X-ray diffraction. Again, this is notdissimilar to what has been found in paint samples.It is possible that black mercury sulphide could beproduced on the photodecomposition of corderoite.Production of calomel would involve reduction ofthe mercury (II) species, with concurrent oxidation,most likely of the S2- ion. The fact that vermilion is asemiconductor facilitates reactivity, especially redoxprocesses. Elemental sulphur could be produced,and, to balance the equation, mercury sulphide,which could be in the black form.

This reaction sequence can be summarised as initialphotosensitisation of vermilion by chloride ions in ahumid environment, followed by degradation by theaction of light. It must be stressed that this specificproposed mechanism is speculative, but a mecha-nism of this type is supported by the fact that it hasbeen shown that the light-induced darkening ofvermilion pigment is faster under conditions ofhigher humidity.

The influence of painting technique

The main agents in the deterioration of vermilionappear to be light, humidity and certain ionic species,which in the paintings examined here has been shownto be chloride. The fact that an external agent such aschloride is involved in the deterioration process isconsistent with the unpredictable behaviour ofvermilion. Further understanding of this behaviourcan be gained from examination of samples fromblackened and unblackened areas on the same paint-ing. The results indicate that painting technique is akey factor in determining whether the vermilion on apainting is vulnerable to discoloration.

56 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23

Marika Spring and Rachel Grout

Hg3S2Cl

2

corderoite

light

Hg2Cl

2

calomelHgS

black cubic metacinnabar

or blackamorphous

mercurysulphide

Ssulphur

+ +

Page 10: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

In Benozzo Gozzoli’s Virgin and Child Enthroned,the vermilion has been protected from discolorationwhere it has been glazed with red lake in the shadowsof the drapery. R.L. Feller has already suggested thata red lake glaze over a vermilion paint will reduce thelevel of light to which it is exposed and filter out lightof certain wavelengths, but it could also act as aphysical barrier to the chloride ions.19 Similarly theshadows of the lining of the cloak of Saint Paul onNiccolò di Pietro Gerini’s Baptism of Christ haveremained red where the vermilion has been protectedby a red lake glaze applied over it. In the discolouredarea vermilion was used alone, and the paint isnoticeably lean and poorly bound. A sample from theundiscoloured lighter orange-red highlights showedthat here the uppermost layer of paint consists of amixture of vermilion and red lead. Samples from theundiscoloured area of Saint Jerome’s robe on thepainting by Giovanni del Ponte were also found toconsist of vermilion mixed with some red lead; inthese cases the red lead appears to have protected thevermilion from discoloration. Paint films containingvermilion are often poorly bound, and it may be thatin a mixture with red lead a more coherent andresilient paint film is formed, in which the pigment isbetter protected from the environment by the bindingmedium.20

Vermilion seems to be particularly vulnerable to

discoloration in paints where it has been used on itsown. On the panel attributed to the Workshop ofthe Master of the Female Half-Lengths, parts ofSaint John’s drapery have discoloured (plate 10). Asample showed that in the discoloured area theuppermost paint layer consists of only vermilionbound in a drying oil. Particles of vermilion havebeen smeared across the resin block during grindingof the sample, suggesting that the pigment is ratherweakly bound. Vermilion is known to accelerate thephoto-oxidation of drying oils, which may explainthe friable nature of the paint.21 Four of the paint-ings that were examined were oil paintings and ourresults were very similar for both these and thetempera paintings. On Jordaens’ Holy Family,another oil painting, the strokes of paint that havediscoloured are extremely thin and are now rathermatt in appearance. In cross-section these strokes

NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 57

The Blackening of Vermilion: An Analytical Study of the Process in Paintings

plate 10 Workshop of the Master of the Female Half-lengths, Saint John on Patmos (NG 717), c.1525–50. Detailof Saint John.

plate 11 Jacob Jordaens, The Holy Family with SaintJohn the Baptist. a. Detail of discoloured vermilion on theVirgin’s cloak.

plate 11 b. Cross-section of a sample from the area ofdiscoloured vermilion (plate 11a). White particles ofmercury chloride, black particles and unchanged red parti-cles of vermilion can be seen in the thin paint layer at thetop of the cross-section. Original magnification 937×;actual magnification 590×.

Page 11: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

can be seen on top of a thick layer of red lake (plate

11). The layer looks very friable and poorly bound,which is probably why the pigment has been vulner-able to the action of chloride ions from theenvironment. The vermilion coat of the horsemanin Aelbert Cuyp’s Hilly River Landscape with aHorseman talking to a Shepherdess (NG 53) hasdiscoloured in certain areas of the modelling. In asample from an undiscoloured area a lead-contain-ing pigment, probably red lead, was detected inaddition to vermilion, and as on tempera paintingsthis has resulted in a more coherent film, protectingthe vermilion from deterioration.22

The influence of conservation history

The conservation history of a painting must also bean important factor in the discoloration of vermil-ion, since it determines whether the painting hasbeen exposed to the environmental conditions thatpromote the colour change. It is clear that the paint-ings in this study have been exposed to chloride atsome point in their history. Most of them have been in the National Gallery Collection since thenineteenth century, and have a reasonably well-documented conservation history from then on. Therecords do not reveal any direct exposure to chlori-nated materials, such as their use for cleaning. Thefact that chlorinated compounds have sometimesbeen used in air-conditioning systems as a disinfec-tant should also be considered but, as mentionedabove, the vermilion in at least one of the paintingsstudied was already discoloured before the generalintroduction of air-conditioning.

However, many of the paintings in this studyseem to have been described as very dirty at somepoint in the conservation records. There is alsoevidence from these accounts, and from examina-tion of the surface coatings, that earlier treatmentsmay have left the paint surfaces of some of thepaintings particularly exposed to contact with dirt.The panels of the San Pier Maggiore Altarpiece, forexample, were cleaned in the 1980s for the first timesince the altarpiece entered the collection in 1857. Athin surface coating of wax polish, applied in 1953,was removed. Underneath was a yellow-grey layer ofdirt, lying directly on the paint surface, suggestingthat when the altarpiece was last cleaned, probablywhile it was in the Lombardi-Baldi collection inFlorence, it was not revarnished. A detail of the leftpanel of the main tier during cleaning (see plate 12)illustrates how heavy this deposit of dirt was.23

Nardo di Cione’s Saint John the Baptist with

Saint John the Evangelist(?) and Saint James (NG581) also entered the National Gallery Collection in1857 and has a similar conservation history. Thepainting was cleaned in the early 1980s for the firsttime since 1887. The varnish that was removedduring this cleaning was extremely thin, and themajority of the discoloration was attributed to greydirt layers that had accumulated on the surface. Incross-sections taken before cleaning in the 1980s,dirt can be seen lying directly on the paint surfacebeneath the very thin varnish applied in 1887,suggesting that it was not thoroughly cleaned atthat time.24

The dirt on these paintings may be responsiblefor the deterioration of the vermilion. As hasalready been mentioned, the high level of dirt in thegalleries in the nineteenth century was a significantand much-discussed problem. Most of the paintingsin the collection were glazed during the second halfof the century to protect them from the sooty air,and remained so until after the Second World War.This measure may not have been entirely effectivehowever, since in 1936, when Ian Rawlins examinedsome pictures that had been glazed, he stated that‘the amount of dust, both on the inside of the glassand on the picture itself, was extraordinarily

58 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23

Marika Spring and Rachel Grout

plate 12 Attributed to Jacopo di Cione and Workshop,San Pier Maggiore Altarpiece. Detail of the left panel ofthe main tier during cleaning.

Page 12: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

large’.25 In any case, the ‘during cleaning’ photo-graphs and cross-sections taken before cleaningprovide evidence that significant levels of dirt werepresent on the paint surfaces from the time theyentered the collection until their cleaning in the1980s.

The very thin varnish applied by Dyer on thealtarpiece by Nardo di Cione was based on mastic.Several of the other paintings in the study have simi-lar thin varnishes, also applied by Dyer in thenineteenth century.26 Towards the end of thecentury, restorers employed by the Gallery werebeing instructed to use the whitest possible spiritvarnish, thinly applied.27 The aim was, presumably,to limit the effects of discoloration. Paintings suchas Benozzo Gozzoli’s Virgin and Child Enthroned,Niccolò di Pietro Gerini’s Baptism of Christ andGiovanni del Ponte’s Ascension of Saint John theEvangelist, with Saints, which still have Dyer’svarnish, are not drastically discoloured – areas ofwhite appear white rather than yellow, despite thefact that over 100 years have passed. However, it isnot clear whether these thin varnishes have in factgiven sufficient protection to the paint surfaces fromthe environment. Thin varnishes allow more light toreach the paint surface; a thick discoloured yellowvarnish can act as a filter, reducing the rate ofdiscoloration, and also acts as a physical barrier todirt.28 It is not known how early the discolorationoccurred on the paintings that have been studied,and records of their conservation history are onlyavailable, at best, for the last century and a half.However, discussion of their history does identifyfactors that could put the paintings at risk to expo-sure to chloride, which can direct future work onpossible preventive conservation measures.

Paintings from the Courtauld Gallery

In all the ten paintings from the National Gallerywith severely discoloured vermilion that were exam-ined, chloride was found to be involved in thedeterioration process. Samples from two paintingsfrom the Courtauld Gallery were also analysed toinvestigate whether the phenomenon extends topaintings in other collections. The paintings arelisted in Table 1, p. 50. Several areas of vermilion onthe central panel of a triptych by Bernardo Daddidepicting the Crucifixion have discoloured. Thedeteriorated areas, such as the headdress of asoldier standing to the right of the cross and therobe of the figure standing behind, have a greyishcolour which is rather similar to the deteriorated

areas on the National Gallery paintings. A cross-section of a sample from a discoloured area of theMagdalen’s robe was examined. Both black andwhite material were visible on the paint surface, asin the National Gallery paintings, and again EDXanalysis detected mercury and chlorine in the whitematerial. A similar process seems therefore to haveoccurred on these paintings. The discoloration inthe Magdalen’s robe follows the modelling, and ithas been found that the areas that are still redcontain red lead rather than vermilion.29

Part of the modelling in the Magdalen’s redcloak on The Estouteville Triptych was painted withvermilion that has now discoloured. EDX analysisestablished that chloride was also involved in thedeterioration on this painting. The painting was in aprivate collection in Sussex as recently as the 1920s,and so was not exposed to the extremely pollutedand dirty air of London in the nineteenth century,suggesting that central London is not the only loca-tion where chloride levels have been high enough toinitiate this deterioration.30

Conclusion

The most important finding of this study is that inall the paintings examined the vermilion is notinherently unstable, but appears to have become soby the action of chloride ions from the environment.The mercury chloride compound formed is not atransient species, as has been suggested in earlierstudies, but the white mercury (I) chloride calomel.The presence of both white and black deteriorationproducts explains why discoloured vermilion oftenappears grey rather than black. Certain paintingsseem to be predisposed to this type of deteriorationas a result of the painting technique employed.Vermilion is vulnerable if used alone in a paint film,which explains why discoloured vermilion is mostcommonly observed on Early Italian paintings,particularly in the pure vermilion mid-tones of thesystematically modelled draperies. The findingssuggest that it is desirable to limit contact betweenvermilion and dirt, or other sources of chlorideions, an aim that has practical consequences forboth preventive and interventive conservation proce-dures.31 Some questions remain to be investigated,such as whether the involvement of chloride in thedeterioration is widespread, the role of varnish inproviding protection against the action of chloride,and whether the level of chloride present in today’sair-conditioned galleries remains sufficient to initi-ate the deterioration of vermilion.

NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 59

The Blackening of Vermilion: An Analytical Study of the Process in Paintings

Page 13: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

Acknowledgements

We are particularly grateful to Janet Ambers(British Museum Research Laboratory) for her helpwith Raman microscopy analysis. Aviva Burnstockkindly allowed us to analyse samples from paintingsfrom the Courtauld Gallery. We would also like tothank John McCormack (Mackay School of Mines,University of Nevada, USA) and Professor DavidVaughan (Manchester University) for useful discus-sion of results.

Notes and references

1 See R.L. Feller, ‘Studies on the Darkening of Vermilionby Light’, Report and Studies in the History of Art,1967, National Gallery of Art, Washington DC, pp.99–111, and V. Daniels, ‘The Blackening of Vermilionby Light’, Recent Advances in the Conservation andAnalysis of Artifacts, Jubilee Conservation ConferencePapers, compiled by James Black, University ofLondon, Institute of Archaeology, 1987, pp. 280–2.

2 R.J. Gettens, R.L. Feller and W.T. Chase, ‘Vermilionand Cinnabar’, Artists’ Pigments: A Handbook ofTheir History and Characteristics, Vol. 2, ed. A. Roy,1993, pp. 159–82. See also Feller (cited in note 1) wherethe work of Eibner (1914) is reviewed. Eibner showedthat wet-process vermilion is less stable than dry-process vermilion, and attributed this to impuritiesintroduced during manufacture, specifically thiosul-phate.

3 Natural mineral cinnabar may sometimes have beenused as a pigment in earlier paintings, although it seemslikely that dry-process vermilion (made by sublimation)was much more common, because of the difficulties ofobtaining cinnabar of a good colour. They are verysimilar microscopically, and no attempt was made todetermine whether the vermilion in the samples exam-ined was cinnabar or the dry-process form. See Gettenset al. cited in note 2.

4 A. Roy and D. Gordon, ‘Uccello’s Battle of SanRomano’, National Gallery Technical Bulletin, 22,2001, p. 10.

5 The Raman spectrum was collected using an excitationwavelength of 784.71 nm.

6 Daniels 1987 (cited in note 1) and R. Grout and A.Burnstock, ‘A study of the blackening of vermilion’,Zeitschrift für Kunsttechnologie und Konservierung,Jahrgang 14/2000, Heft 1, pp. 15–22.

7 The Raman scattering in the spectrum of a standardsample of black metacinnabar is probably beingproduced from traces of hexagonal cinnabar, withcubic black metacinnabar showing no scattering.Metacinnabar should, theoretically, give an entirelydifferent spectrum from vermilion because of its differ-ent crystal structure. See J.R. Ferraro, ‘Factor GroupAnalysis for Some Common Minerals’, Applied

Spectroscopy, 29, No.5, 1975, pp. 418–21.8 Quartz (SiO

2) is often found as an accessory mineral

with cinnabar, but in these samples no discrete colour-less particles of quartz are visible, suggesting that thesilicon is contained within the structure of the mercurysulphide lattice.

9 D. Saunders, ‘Pollution and the National Gallery’,National Gallery Technical Bulletin, 21, 2000, p. 90 andp. 93, note 83.

10 K. Gysels and R. Van Grieken, ‘Microanalysis ofMuseum Aerosols related to the conservation of worksof art’, Proceedings of the 6th InternationalConference on ‘Non-destructive Testing andMicroanalysis for the Diagnostics and Conservation ofthe Cultural and Environmental Heritage’, Rome, May1999, Vol.1, pp. 45–62.

11 The EDX analysis was not quantitative. Some attemptwas made to characterise the different particles makingup the dirt by a series of spot analyses. Discrete parti-cles of sodium chloride could be detected, as well ascalcium-rich particles, calcium silicates and particlescontaining a combination of Ca, S, Cl and Si.

12 The blackening of the vermilion on Uccello’s Battle ofSan Romano can be seen in a colour plate dating fromthe 1930s. See R.N.D. Wilson, The National and TateGalleries, London 1930s (exact date of publication notknown), plate 2. Air-conditioning was introduced toone room in the National Gallery in September 1950,followed by another six rooms in 1956. See TheNational Gallery, 1938–54, London 1955, pp. 23–4.Measurements made in 1959 showed that the air-condi-tioning system removed over 90% by weight of dust.The level in a naturally ventilated gallery was around80% of that outside. The clean air act of 1956 also ledto a significant reduction in the level of pollution incentral London. See Saunders 2000, cited in note 9.

13 P. Brimblecombe, Air Composition and Chemistry, 2ndedn, Cambridge 1995, p. 110.

14 P. Brimblecombe, The Big Smoke. A history of airpollution in London since medieval times, Cambridge1987, p. 146.

15 Daniels 1987, cited in note 1, refers to a number ofpapers by Davidson and Willsher, and Davidson et al.in which potassium iodide, chloride, bromide and fluo-ride, ferrocyanide, iodate, thiocyanate, cyanide andlead nitrate were found to accelerate the blackening ofvermilion.

16 Daniels, cited in note 1.17 J.K. McCormack, ‘The darkening of cinnabar in

sunlight’, Mineralium Deposita, 35, 2000, pp. 796–8.18 J.K. McCormack, personal communication.19 Feller 1967, cited in note 1.20 Lead could also react preferentially with chloride ions,

thereby protecting the vermilion.21 F. Rasti and G. Scott, ‘The effects of some common

pigments on the photo-oxidation of linseed oil-basedpaint media’, Studies in Conservation, 25, 1980, pp.145–56.

22 J. Plesters, A. Roy and D. Bomford, ‘Interpretation ofthe magnified image of paint surfaces and samples in

60 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23

Marika Spring and Rachel Grout

Page 14: National Gallery Technical Bulletin · NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 51 The Blackening of Vermilion: An Analytical Study of the Process in Paintings plate 1 Jacob

terms of condition and appearance of the picture’,Science and Technology in the Service of Conservation,Preprints of the IIC Washington Congress, September1982, N.S. Bromelle and G. Thomson eds, London1982, pp. 169–76.

23 J. Dunkerton, J. Kirby and R. White, ‘Varnish and earlyItalian tempera paintings’, Cleaning, retouching andcoatings, Preprints of the contributions to the BrusselsIIC Congress, September 1990, J.S. Mills and P. Smith,eds, London 1990, pp. 63–9. This paper discusses thedocuments that survive which record the varnishing ofthe San Pier Maggiore Altarpiece soon after it wascompleted. Traces of an oil/resin varnish (linseed oilplus sandarac) were found on one of the panels, whichmay be the remains of the original varnish. This type ofvarnish is likely to have provided good protection for thepaint surface. It is not known when it was removed, butthe painting appears to have been unvarnished when itentered the National Gallery Collection in 1857. It haslong been the practice of restorers to apply the mini-mum amount of varnish to a painting of this periodafter restoration, to produce a surface of low gloss.

24 D. Gordon, D. Bomford, J. Plesters and A. Roy, ‘Nardodi Cione’s “Altarpiece: Three Saints”’, National GalleryTechnical Bulletin, 9, 1985, pp. 21–37.

25 Saunders 2000, cited in note 9. 26 R. White and J. Kirby, ‘A Survey of Nineteenth- and

early Twentieth-Century Varnish Compositions foundon a selection of Paintings in the National GalleryCollection’, National Gallery Technical Bulletin, 22,2001, pp. 64–84. Several of the paintings discussed inthis paper had varnishes applied by Dyer.

27 An Exhibition of Cleaned Pictures (1936–1947), TheNational Gallery, London 1947, p. xvii.

28 Feller (cited in note 1) found that a yellow filter reducedthe rate of discoloration of vermilion, although not aseffectively as a red filter.

29 The two paintings in this study from the CourtauldGallery are illustrated in Grout and Burnstock, cited innote 6.

30 Catalogue of the Lee Collection, Courtauld Institute ofArt, University of London, revised edn 1962, p. 10.

31 A factor that needs to be taken into considerationduring conservation treatment of paintings withseverely discoloured vermilion is that the whitecalomel-containing material often appears to be poorlybonded to the surface of the unchanged vermilion (asseen in the cross-section from the painting by Niccolòdi Pietro Gerini (plate 8c)) and may be vulnerable toremoval during cleaning. It may be the reason for thevariation in the proportion of black and white materialthat was observed in the deteriorated crust. It should,perhaps, also be taken into consideration that whencalomel is treated with ammonia, disproportionationcan occur, producing a black solid of elementalmercury and a complex mixture of mercury(II)compounds. However, whether this is a problem inpractice, given that aqueous ammonia solution isusually only used in low concentration during conser-vation treatments, requires further investigations. See

N.N. Greenwood and A. Earnshaw, Chemistry of theElements, Oxford 1984 (1990 reprint), p. 1410.

NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 61

The Blackening of Vermilion: An Analytical Study of the Process in Paintings